Chemical Physics Letters 397 (2004) 51–55
www.elsevier.com/locate/cplett
Confirmation of enhanced anion concentration
at the liquid water surface
Poul B. Petersen, Richard J. Saykally
*
Department of Chemistry, University of California, B84 Hildebrand Hall, Berkeley, CA 94720-1460, USA
Received 13 July 2004; in final form 13 July 2004
Available online 12 September 2004
Abstract
The textbook view of the liquid electrolyte surfaces as being devoid of ions have recently been challenged by molecular dynamics
simulations, which predict a surface enhancement of highly polarizable anions. Here we present the first direct experimental verification of this prediction. Enhanced azide ðN
3 Þ concentrations were measured at the liquid surface by femtosecond second harmonic generation (SHG) experiments exploiting the charge-transfer-to-solvent (CTTS) resonance of N
3 , yielding a surface excess
free energy of 9.9 ± 0.3 kJ/mole. Such surface-enhanced concentrations of anions could have important consequences for the
chemical reactions taking place on atmospheric aerosols and at the ocean-air interface.
2004 Elsevier B.V. All rights reserved.
1. Introduction
Theoretical models of electrolyte solutions have until
recently been limited to a description of the solvent as a
continuum dielectric medium and the interface as a discontinuous change in dielectric constant. The earliest
model, by Onsager and Samaras [1] described the ions
as point charges and employed image charge repulsion
as the dominant interfacial force, as first suggested by
Wagner [2] Since then, several improvements to the original model have been made, taking into account finite
ion size, dispersion forces, hydrophobicity and polarizability, as described in recent review articles [3,4]. These
models extend the agreement with surface tension experiments, but incorporate empirical and even unphysical
parameters, such as a cut-off distance of closest approach even for ions that are attracted to the surface.
Moreover, they cannot describe the molecular interac-
*
Corresponding author. Fax: +1 510 642 8566.
E-mail address: [email protected] (R.J. Saykally).
0009-2614/$ - see front matter 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.cplett.2004.08.049
tions between ions and water molecules that account
for the different solvation of negative and positive
charges in water, or the molecular-level structure of
the liquid–vapor interface. An alternative approach is
to use MD simulations that explicitly incorporate
molecular water–ion interactions and molecular surface
structure, at the cost of longer calculation time and loss
of analytical expressions. These simulations are very
sensitive to the potentials used. Those employing nonpolarizable potentials [5, 6] generally predict all ions to
be repelled from the surface, whereas those using polarizable potentials [7, 8, 9] predict that large polarizable
anions are actually attracted to the outermost surface
layers, while small unpolarizable ions are repelled from
the surface, as in the early models. This is consistent
with both experiments and theoretical calculations on
aqueous clusters, which show the polarizable anions to
reside at the surface, whereas the cations and non-polarizable anions remain in the interior of the clusters
[10,11].
Experimental investigations have similarly been limited to macroscopic thermodynamic methods such as
52
P.B. Petersen, R.J. Saykally / Chemical Physics Letters 397 (2004) 51–55
surface tension and surface potential measurements. The
surface tension of electrolyte solutions increases linearly
with concentration above 0.01 M [12]. A surface tension
increase is linked to a surface deficit by the Gibbs
adsorption equation [13]. Linear optical methods have
only been applicable for surfactants rigorously confined
to the surface, as light necessarily penetrates at least
wavelength dimensions into the solution, probing hundreds of molecular layers. Indirect experiments have alluded to the presence of ions at the interface by
measuring chemical reaction dynamics that could only
be explained by reactions involving surface ions
[14,15]. Recent technological advances have made it possible to directly and selectively probe the liquid water
surface and soluble surfactants on a molecular level
using nonlinear optical and photoelectron techniques.
Recent studies by the Richmond [16] and Allen [17]
groups have addressed the local water structure of the
interface of halide salt solutions in the 0.5–2 M concentration range by the very similar surface selective
technique of sum-frequency generation (SFG) spectroscopy. Their results indicate only a minor perturbation of
the water structure at the interface, but reveal an increased interfacial thickness for bromide and iodide
solutions. However, these indirect experiments were
not able to establish or exclude the existence of an enhanced ion concentration in the outermost surface layers. Photoelectron emission experiments have been
able to directly detect iodide at the interface in the presence of large hydrophobic counter-ions [18] and concentrations of sodium iodide solutions higher than 0.1 M
[19]. These results show surface saturation above 2 moles/kg, resembling a traditional Langmuir isotherm,
which indicates an enhanced surface concentration.
However, the authors conclude that the surface is instead depleted of ions, in agreement with the Gibbs
adsorption equation, but they were unable to exclude
an enhancement in the outmost surface layers because
the probing depth of their experiment (several nm), is
much larger than the thickness of a water layer (about
3 Å). The recent theoretic and experimental findings
have received much attention, engendering debate over
possible violation of the Gibbs adsorption equation
[20–22].
In this Letter, we report verification of the predicted
surface anion enhancement in the outermost molecular
layers of aqueous solutions of sodium azide through
the use of surface-specific and molecule-selective resonant second harmonic generation (SHG) spectroscopy.
The measured concentration profiles yield a Langmuir
adsorption isotherm for surface versus bulk azide concentrations, from which the corresponding negative
surface free energy (DGads = 9.9 ± 0.3 kJ/mol) is extracted. The azide anion has previously been investigated both in the gas phase [23] and the bulk liquid,
showing a charge-transfer-to-solvent (CTTS) band
Fig. 1. Spectral response. Data points are the magnitude of the azide
response defined as the norm square of the complex susceptibility,
(B 0 )2 + (C 0 )2, obtained from the fit in Fig. 2. Error bars are calculated
from the statistical uncertain obtained in the fit. The solid line is the
bulk absorbance of sodium azide, showing the bulk CTTS band.
around 200 nm [24]. A recent paper characterizes small
azide-containing water clusters by both photoelectron
spectroscopy and MD simulations as well as azide at
the extended water surface by MD simulations [9]. In
all cases, they find the azide anion strongly adsorbing
to the surface. The extended surface simulation was carried out with 1 azide anion and 556 water molecules,
corresponding to a 0.1 M solution. Using the surface
specific technique of SHG, we directly probe the azide
adsorption through the resonance structure of the surface CTTS band at three different wavelengths: 200,
225 and 250 nm. The bulk CTTS band is shown in
Fig. 1.
2. Experimental
The laser system will be described in detail in a future
publication and only a brief description is presented
here. A homebuilt Ti:sapphire oscillator is used to seed
a regenerative amplifier (Spectra Physics, Spitfire). The
amplifier is used to pump two OPAÕs (Light Conversion,
TOPAS). The three probing wavelengths used in this
Letter are generated by doubling the 800 nm directly
(400 nm), by fourth harmonic generation of the idler
from the TOPAS (450 nm) and by sum-frequency of
the 800 nm with the signal beam from the TOPAS
(500 nm).
All glassware in contact with the solutions are soaked
in hot chromic acid and rinsed thoroughly with 18 MX
water (Millipore, Milli-Q) before use. The sodium azide
solutions were prepared with 18 MX water and 99.99%
pure sodium azide obtained from Aldrich and used without further purification.
P.B. Petersen, R.J. Saykally / Chemical Physics Letters 397 (2004) 51–55
is incorporated in the expression for the SHG intensity
assuming the water background to be constant and the
orientation of the azide ions stay constant
3. Results and discussion
Like other even-order nonlinear processes, SHG is
forbidden in bulk centro-symmetric media within the dipole approximation [25]. Bulk liquid water exhibits
inversion symmetry due to the randomized orientation
of the water molecules, but at the liquid–air interface,
the symmetry is necessarily broken and the top few
molecular layers exhibit a net orientation due to the
interface and generate a weak SHG response. For the
pure air–water interface, the majority of the SHG signal
is generated in the top liquid layer with a small contribution from the second layer, according to MD simulations [26]. SHG, as well as the similar technique of
SFG, have become a widely used surface-selective probe
for liquid surfaces [27].
The SHG intensity is expressed by the second order
susceptibility v(2), that in turn is the sum of the contributions from the water background and the azide anions:
2
I SHG /j vð2Þ j I 2Fundamental ;
ð2Þ
I SHG /jA þ ðB þ iCÞ N S j
ð2Þ
The susceptibilities are in general complex quantities
but non-resonant contributions, such as the water background, are real. The norm square of the total susceptibility in Eq. (1) leads to interference between the
different contributions. Both constructive and destructive interference with the water background have previously been observed [28]. The azide susceptibility is
proportional to the surface concentration, NS, and the
orientationally averaged molecular susceptibility (hyperpolarizability, b)
ð2Þ
ð3Þ
Assuming the orientation of the azide molecules does
not change significantly over the concentration range,
the change in the azide response is from surface concentration changes.
The surface adsorption is modeled by the standard
Langmuir adsorption model
NS ¼
N max
x
S
:
x þ 55:5 moles=liter expðDGAds =RTÞ
ð5Þ
Here A is the real water background, B and C are
the real and imaginary component of the azide susceptibility and D is the Langmuir constant 55.5
M · exp(DGAds/RT). The constants A, B 0 and C 0 are allowed to change with the wavelength but the constant
D is fitted simultaneously to all three wavelengths.
The magnitude of the azide susceptibility (B 0 )2 + (C 0 )2,
for the different wavelengths are shown in Fig. 1. The
Gibbs free energy of adsorption is extracted from the
constant D. The obtained fit parameters are shown in
Table 1.
The SHG intensity measured as a function of concentration for the three selected wavelengths is shown in
Fig. 2. The SHG intensity is observed to increase monotonically with azide concentration at the resonant wavelength 200 nm. At the near-resonant wavelengths, 225
and 250 nm, the SHG intensity shows a small initial decrease due to the partly destructive interference between
the azide resonance and background water response
[28]. The concentration dependence at the three wavelengths is fit simultaneously to the standard Langmuir
adsorption equation, as described in the methods section. The fit yields a Gibbs free energy of adsorption
of 9.9 ± 0.3 kJ/mole, corresponding roughly to the energy of half a water hydrogen bond. The rest of the fitting parameters are shown in Table 1. The negative
value of the free energy is in accord with surface
enhancement of the azide anion at the air–water interface predicted by Joungwirth [9]. However, their theoretical prediction does not include the Gibbs free
energy.
We believe this is the first direct measurement of enhanced ion concentration at the liquid water–air interface. It should be noted that due to the very shallow
probing depth of the SHG experiment [26]. the measured Gibbs free energy is for adsorption to the outermost molecular layers of liquid density. Different
surface probes have different probing depths and
ð2Þ
vazide ¼ N S hbiOrientation :
2
¼ ðA þ B N S Þ2 þ ðC N S Þ2
2 0
2
B0 x
C x
¼ Aþ
þ
:
xþD
xþD
ð1Þ
vð2Þ ¼ vwater þ vazide :
53
ð4Þ
is
Here x is the bulk electrolyte concentration, N max
S
the maximum surface concentration and DGAds is the
Gibbs free energy of adsorption. The Langmuir model
Table 1
Parameters obtained in the fit of the SHG data to the Langmuir model in Fig. 2
Wavelength (nm)
A
B0
C0
(B 0 )2 + (C 0 )2
200
225
250
0.988 ± 0.006
0.982 ± 0.006
0.991 ± 0.006
0.30 ± 0.27
0.82 ± 0.12
0.58 ± 0.12
3.66 ± 0.12
2.84 ± 0.16
1.87 ± 0.14
13.5 ± 1.9
8.7 ± 1.7
3.8 ± 1.3
The constant D is fitted globally to yield the value 0.92 ± 0.12.
54
P.B. Petersen, R.J. Saykally / Chemical Physics Letters 397 (2004) 51–55
4. Conclusion
Using femtosecond second harmonic generation we
have directly measured an enhanced azide ðN 3 Þ concentration at the liquid water–air interface as predicted by
recent computer simulations [9]. The concentration profile yield a Langmuir isotherm with a Gibbs free energy
of adsorption of 9.9 ± 0.3 kJ/mole.
These recent findings highlight the need for new theoretical models of the liquid–vapour interfaces of aqueous electrolytes, explicitly incorporating the molecular
structure of the solvent that engenders the specific solvation of negative and positive charges and the local structure of the water surface, such as those obtained in MD
simulations. For these, the effective potentials used are
of central importance and reliable direct experimental
measurements, such as those presented here, are essential to develop realistic models.
Acknowledgements
This work is funded by the Experimental Physical
Chemistry Division at the National Science Foundation.
P.B.P. is funded by the Danish Research Training
Council.
References
Fig. 2. SHG intensity and the extracted azide surface concentration at
wavelengths 200, 225 and 250 nm. The SHG intensity is normalized to
pure water. Panel A shows the SHG response and Panel B shows the
surface concentration as a function of bulk concentration. The solid
black squares, red circles and green triangles are experimental data at
200, 225 and 250 nm, respectively. Error bars are estimated from the
reproducibility of two data series. The solid lines are fit to a SHG
Langmuir response as described in the text. The data sets are fit
simultaneously to yield a Gibbs excess free energy of adsorption of
9.9 ± 0.3 kJ/mole. The main panels show concentrations on a
logarithmic scale, where as the insets show the same data on a linear
scale.
accordingly measure different surface properties. The
distribution of ions at the interface could also be highly
non-monotonic along the surface normal, as predicted
for bromide and iodide, wherein the surface enhancement at the outermost layer is followed by a depletion
in subsequent layers [7]. In these cases, the surface concentration integrated through the entire interfacial region should be lower than the bulk concentration, in
agreement with the Gibbs adsorption equation and
the observed surface tension increase, even though
the top surface layer is enhanced.
[1] L. Onsager, N.N.T. Samaras, J. Phys. Chem. 2 (1934) 528–536.
[2] V.C. Wagner, Physik Zeirschr. 15 (1924) 474.
[3] M. Manciu, E. Ruckenstein, Adv. Colloid Interfac. Sci. 105 (2003)
63.
[4] K.A. Karraker, C.J. Radke, Adv. Colloid Interfac. Sci. 96 (2002)
231.
[5] S. Paul, A. Chandra, Chem. Phys. Lett. 373 (2003) 87.
[6] D. Bhatt, J. Newman, C.J. Radke, J. Phys. Chem. B 108 (2004)
13412.
[7] P. Jungwirth, D.J. Tobias, J. Phys. Chem. B 106 (2002) 6361.
[8] P. Salvodor, J.E. Curtis, D.J. Tobias, P. Jungwirth, Phys. Chem.
Chem. Phys. 5 (2003) 3752.
[9] X. Yang, et. al., J. Phys. Chem. A, ASAP article, web release date:
May 8th (2004).
[10] L. Lehr, M.T. Zanni, C. Frischkorn, R. Weinkauf, D.M.
Neumark, Science 284 (1999) 635.
[11] W.H. Robertson, M.A. Johnson, Annu. Rev. Phys. Chem. 54
(2003) 173.
[12] P.K. Weissenborn, R.J. Pugh, J. Colloid Interf. Sci. 184 (1996)
500.
[13] A.W. Adamson, A.P. Gast, Physical Chemistry of Surfaces, sixth
ed., Wiley, New York, 1997.
[14] J.H. Hu et al., J. Phys. Chem. 99 (1995) 8768.
[15] I. Watanabe, N. Takahashi, H. Tanida, Chem. Phys. Lett. 287
(1998) 714.
[16] E.A. Raymond, G.L. Richmond, J. Phys. Chem. B 108 (2004)
5051.
[17] D. Liu, G. Ma, L.M. Levering, H.C. Allen, J. Phys. Chem. B 108
(2004) 2252.
[18] J. Kohno, F. Mafune, T. Kondow, J. Phys. Chem. A 105 (2001)
5990.
[19] R. Weber et al., J. Phys. Chem. B 108 (2004) 4729.
P.B. Petersen, R.J. Saykally / Chemical Physics Letters 397 (2004) 51–55
[20] B.C. Garrett, Science 303 (2004) 1146.
[21] D. Bhatt, R. Chee, J. Newman, C.J. Radke, Curr. Opin. Colloid
In., available online 23 July 2004.
[22] L. Vrbka et. al., Curr. Opin. Colloid In., available online 3 July
2004.
[23] M. Polak, M. Gruebele, R.J. Saykally, J. Am. Chem. Soc. 109
(1987) 2884.
55
[24] I. Burak, A. Treinin, J. Chem. Phys. 39 (1963) 189.
[25] Y.R. Shen, The Principles of Nonlinear Optics, Wiley, New York,
1984.
[26] V.P. Sokhan, D.J. Tildesley, Mol. Phys. 92 (1997) 625.
[27] K.B. Eisenthal, Chem. Rev. 96 (1996) 1343.
[28] G.A. Reider, M. Cernusca, M. Hofer, Appl. Phys. B 68 (1999)
343.